Not yet, if for no other reason, then because it seems too volatile. I wouldn't want to take the time to write a big project in Arc right now, when future releases of Arc will be incompatible with my code in possibly entirely unforeseeable ways.
A module system would be a big plus, too, though I (and many other people) have successfully managed rather large C projects sans a real module system.
Not sure. Could be because the wordier approach more clearly states what is going on, and makes it easier to understand anonymous lambda expressions later on?
If pg & co. will add the comments and/or accept patches from users with comments, then yes, this would be the best method for function reference documentation, especially given Arc's present volatility.
Surely more documentation is forthcoming from pg when he has time, and some other folks will likely write up docs until then. I started on "redocumenting" the arc spec file, but I don't have oodles of free time myself... A wiki-esque page would be nice.
Somewhere else on this forum, the point was made that the Arc webapp code works properly on BSD-based systems, but not GNU/Linux systems, due to an incompatibility in time representation.